Application of neural networks for classification of biological compounds from the characteristics of whispering-gallery-mode optical resonance

We have analyzed the feasibility and expediency of using a probabilistic neural network as a classifier for processing and interpreting data obtained using a biosensor we have developed that is based on whispering-gallery-mode optical resonance. We have established that the probabilistic neural network makes it possible to correctly classify the biological compounds under study with average 97.3% probability.     

Transfer-Learning-Based Approach for the Diagnosis of Lung Diseases from Chest X-ray Images

Using chest X-ray images is one of the least expensive and easiest ways to diagnose patients who suffer from lung diseases such as pneumonia and bronchitis. Inspired by existing work, a deep learning model is proposed to classify chest X-ray images into 14 lung-related pathological conditions. However, small datasets are not sufficient to train the deep learning model. Two methods were used to tackle this: (1) transfer learning based on two pretrained neural networks, DenseNet and ResNet, was employed; (2) data were preprocessed, including checking data leakage, handling class imbalance, and performing data augmentation, before feeding the neural network. The proposed model was evaluated according to the classification accuracy and receiver operating characteristic (ROC) curves, as well as visualized by class activation maps. DenseNet121 and ResNet50 were used in the simulations, and the results showed that the model trained by DenseNet121 had better accuracy than that trained by ResNet50.     

Sand/cement ratio evaluation on mortar using neural networks and ultrasonic transmission inspection

The quality and degradation state of building materials can be determined by nondestructive testing (NDT). These materials are composed of a cementitious matrix and particles or fragments of aggregates. Sand/cement ratio (s/c) provides the final material quality; however, the sand content can mask the matrix properties in a nondestructive measurement. Therefore, s/c ratio estimation is needed in nondestructive characterization of cementitious materials. In this study, a methodology to classify the sand content in mortar is presented. The methodology is based on ultrasonic transmission inspection, data reduction, and features extraction by principal components analysis (PCA), and neural network classification. This evaluation is carried out with several mortar samples, which were made while taking into account different cement types and s/c ratios. The estimated s/c ratio is determined by ultrasonic spectral attenuation with three different broadband transducers (0.5, 1, and 2 MHz). Statistical PCA to reduce the dimension of the captured traces has been applied. Feed-forward neural networks (NNs) are trained using principal components (PCs) and their outputs are used to display the estimated s/c ratios in false color images, showing the s/c ratio distribution of the mortar samples.     

Speech segment durations and quantity in Icelandic.

Icelandic has a phonologic contrast of quantity, distinguishing long and short vowels and consonants. Perceptual studies have shown that a major cue for quantity in perception is relational, involving the vowel-to-rhyme ratio. This cue is approximately invariant under transformations of rate, thus yielding a higher-order invariant for the perception of quantity in Icelandic. Recently it has, however, been shown that vowel spectra can also influence the perception of quantity. This holds for vowels which have different spectra in their long and short varieties. This finding raises the question of whether the durational contrast is less well articulated in those cases where vowel spectra provide another cue for quantity. To test this possibility, production measurements were carried out on vowels and consonants in words which were spoken by a number of speakers at different utterance rates in two experiments. A simple neural network was then trained on the production measurements. Using the network to classify the training stimuli shows that the durational distinctions between long and short phonemes are as clearly articulated whether or not there is a secondary, spectral, cue to quantity.     

HL-2A tokamak disruption forecasting based on an artificial neural network

Artificial neural networks are trained to forecast the plasma disruption in HL-2A tokamak. Optimized network architecture is obtained. Saliency analysis is made to assess the relative importance of different diagnostic signals as network input. The trained networks can successfully detect the disruptive pulses of HL-2A tokamak. The results obtained show the possibility of developing a neural network predictor that intervenes well in advance for avoiding plasma disruption or mitigating its effects.     

Hybrid neural nets with poisson and gaussian connectivities

The dynamic behavior of neural nets with different patterns of interneuronal synaptic connectivity is investigated. Our method is based on probabilistic neural nets for the net structure and dynamics. Each net is divided into several different subsystems, which are characterized by different distribution laws for the number of connections that the neurons make. We start from the binomial distribution, which, under appropriate conditions, reduces to the Poisson and Gaussian distributions. The overall net now acquires a hybrid character. The expression for the neural activity is generalized to include this effect, and new expressions are derived, based on the isolated single-net equations. The dynamics of nets with sustained external inputs is also studied. The results obtained by this approach also show multiple stability and multiple hysteresis effects, as in the case of single nets. The differences between pure Poisson, Gaussian, and hybrid nets are explained in terms of the structural properties of the model. As expected, the hybrid case falls in between the two other distributions. Finally, we performed Monte Carlo computer calculations for the hybrid nets. For the range of parameters examined we find very good agreement with the developed formalism     

Artificial neural networks applied to the classification of hair samples according to pigment and sex using non-invasive analytical techniques

In this study, we investigated the possibility of using an artificial neural network (ANN) to classify human hair samples according to pigment (original or bleached hair) and sex (female or male) from numerical data obtained by wavelength dispersive X-ray fluorescence (WDXRF) and by laser-induced breakdown spectroscopy (LIBS). The results were promising, showing that the developed ANNs are able to classify the pigment and donor sex of hair samples with 100% and 89.5% accuracy, respectively, in the test set using WDXRF data. For the LIBS data in the test set, 100% of the pigment classifications were correct, and 78.9% of the donor sex classifications were correct.     

Shape and Size Determination by Laser Diffraction: Feasibility of Data Analysis by Neural Networks

The feasibility of the inversion of laser diffraction data for size and shape distribution by neural networks has been investigated by computer simulation. Neural networks trained with diffraction patterns of elliptical particles with different sizes and aspect ratios (axis ratios) were able to recover simultaneously both the size and aspect ratio distributions in a few milliseconds on a common PC.     

基于级联人工神经网络的生物表面光谱识别方法

提出了一种利用多级级联人工神经网络对生物表面微区的可见光光谱进行识别与分类的方法。该方法利用自组装光纤探头式光谱仪对苹果表面微区500～730 nm范围内的可见光光谱进行测量,光谱间隔5 nm, 记录光谱测量数据并依据光谱测量数据建立由三个单隐层、四十七个输入、单输出的人工神经网络级联而成的光谱识别系统。实验表明该级联系统可以对苹果的烂痕、疤痕、碰痕的反射光谱进行准确识别,在5％和15％的噪声影响下其识别准确率分别能达到97％和85％以上,克服了单级人工神经网络识别准确率不高、抗噪声能力差等缺点。最后文章提出了一种识别结果的隶属度表示法,该方法借鉴模糊数学中隶属度的概念,可以实现对识别结果客观、准确的表征。     

一种基于卷积神经网络的恒星光谱快速分类法

恒星光谱数据的分类是天体光谱自动识别的最基本任务之一,光谱分类的研究能够为恒星的演化提供线索。随着科技的发展,天文数据也向大数据时代迈进,需要处理的恒星光谱数量越来越多,如何对其进行自动而精准地分类成为了天文学家要解决的难题之一。当前恒星光谱自动分类问题的解决方法相对较少,为此本文使用了一种基于卷积神经网络的方法对恒星光谱MK系统进行分类。该网络由数据输入层、四个卷积层、四个池化层、全连接层、输出层构成,与传统网络相比具有局部感知、参数共享等优点实验。在Python3.5的环境下编程,利用Tensorflow构建了一个简单高效的具有四个卷积层的卷积神经网络,并将Dropout作用于全连接层之后以防止过度拟合。Dropout的基本思想：当网络模型进行训练时,把一些神经网络节点按一定的比例丢弃,使其暂时不发挥作用。Dropout可以理解成是一种十分高效的神经网络模型平均方法,由于它不依赖于某些局部特征所以能够让网络模型更加鲁棒。实验中使用的一维恒星光谱图是取自LAMOST DR3数据库,首先进行预处理截取光谱3 600~7 300 Å的部分,均匀采样后使用min-max标准化法对其进行初始化。实验包括两部分：第一部分为依据恒星光谱MK系统对光谱进行分类,每一类的训练样本包含1 000条光谱数据,测试样本为400条光谱数据,首先通过训练样本对CNN网络进行训练,进行3 000次的迭代,用训练后的网络将测试样本进行分类以验证网络的准确性;第二部分为相邻两类的恒星光谱的分类,其中O型星数据集样本为250条光谱,其余类别恒星样本数据集均为4 000条光谱,将数据5等分,每次选取当中的一份当作测试集,其余部分当作训练集,采用5折交叉验证法求得模型准确率,用BP神经网络进行对比实验。选择对网络模型进行评估的指标包括精确率P、召回率R、F-score、准确率A。实验结果显示CNN在对六类恒星光谱进行分类时其准确率都在95%以上,在对相邻类别的恒星进行分类时,由于O型星样本量较少,所以得到的分类结果不太理想,对其余类别的恒星分类准确率都高于98%,以上结果都证明了CNN算法能够很好地解决恒星光谱的分类问题。     

Neural network-based segmentation of dynamic MR mammographic images

The usefulness of neural networks for the classification of signal-time curves from dynamic MR mammography was recently demonstrated by our group. The multi-layer perceptron under study consists of 28 input, 4 hidden, and 3 output nodes, and was trained to classify signal-time curves into three tissue classes: "carcinoma," "benign lesion," and "parenchyma." Extending this approach, it was the aim of the present study to evaluate the performance of the developed network in the segmentation of dynamic MR mammographic images in comparison to a pixel-by-pixel two-compartment pharmacokinetic analysis. The population investigated in this pilot study comprised 15 women with suspicious lesions in the breast, which were confirmed histologically after the MR examination. The neural network classified the same areas as malignant as those which were marked as being highly suspicious by the pharmacokinetic mapping approach but with the advantage that no a priori knowledge on tissue microcirculation was needed, that computation proved to be much faster, and that it yielded a unique classification into just three tissue classes.     

结合小波融合和深度学习的脑胶质瘤自动分割

针对水肿区域边界模糊和瘤内结构复杂多变导致的脑胶质瘤分割不精确问题,本文提出了一种基于小波融合和3D-UNet网络的脑胶质瘤磁共振图像自动分割算法.首先,对脑胶质瘤磁共振图像的T1、T1ce、T2、Flair四种模态进行小波融合以及偏置场校正;然后,提取待分类的图像块;再利用提取的图像块训练3D-UNet网络以对图像块中的像素进行分类;最后加载损失率较小的网络模型进行分割,并采用基于连通区域的轮廓提取方法,以降低假阳性率.对57组Brats2018（Brain Tumor Segmentation 2018）磁共振图像测试集进行分割的结果显示,肿瘤的整体、核心和水肿部分的平均分割准确率（DSC）分别达到90.64%、80.74%和86.37%,这表明该算法分割脑胶质瘤准确率较高,与金标准相近.相比多模态图像融合前,该算法在减少输入网络数据量和图像冗余信息的同时,还一定程度上解决了胶质瘤边界模糊、分割不精确的问题,提高了分割的准确度和鲁棒性.     

On the Optical Diagnosis and Sizing of Spherical Coated and Multilayered Particles with phase-Doppler anemometry

After having recalled basic theoretical results concerning the extension of generalized Lorenz-Mie theory to the case of multilayered spheres, results connected with phase-Doppler anemometry are considered, showing the influence of Gaussian beam intensity profiles on the light scattering properties of these particles. Particular emphasis is placed on the case of water-coated carbon core particles, for which the possibility of obtaining simultaneous size measurements of the core and outer diameters is discussed. The sensitivity of the technique to particles with a refractive index profile is also considered, showing that this technique is more sensitive to changes in the average refractive index of the particles than to refractive index profiles, such as produced by high pressure and temperature stresses, at least for the studied geometry.     

Identification and classification of organic and inorganic components of particulate matter via Raman spectroscopy and chemometric approaches

In this paper, a novel method for developing a tree‐like classifier which differentiates between organic and inorganic particulate matter by means of Raman spectroscopy is introduced. The algorithm is fully automatic and optimises itself without any human interaction. This method uses a tree‐like structure to classify Raman spectra as a decision tree. On every knot of this tree, the optimal classifier is automatically obtained, tested and trained. The optimal classifier is an artificial neural network, linear discriminant analysis or a support vector machine, where different kernels are possible. The support vector machine is optimised by the simulated annealing method to achieve the best possible classifier. After the training, a hold‐out experiment with two completely independent sets of Raman spectra was tried to show the abilities of this method for real‐world application. Copyright © 2010 John Wiley & Sons, Ltd.     

Classification of Two-Dimensional Local Conformal Nets with <Emphasis Type="Italic">c</Emphasis> < 1 and 2-Cohomology Vanishing for Tensor Categories

We classify two-dimensional local conformal nets with parity symmetry and central charge less than 1, up to isomorphism. The maximal ones are in a bijective correspondence with the pairs of A-D-E Dynkin diagrams with the difference of their Coxeter numbers equal to 1. In our previous classification of one-dimensional local conformal nets, Dynkin diagrams D _{2n +1} and E ₇ do not appear, but now they do appear in this classification of two-dimensional local conformal nets. Such nets are also characterized as two-dimensional local conformal nets with -index equal to 1 and central charge less than 1. Our main tool, in addition to our previous classification results for one-dimensional nets, is 2-cohomology vanishing for certain tensor categories related to the Virasoro tensor categories with central charge less than 1.Supported in part by JSPS.Supported in part by GNAMPA and MIUR.     

基于PCA-BP神经网络的光谱消光法颗粒粒径反演算法研究

光谱消光法广泛应用于颗粒粒径测量领域,在利用光谱消光法对颗粒粒径进行反演的过程中,由于颗粒的消光系数存在理论复杂、计算繁琐、收敛速度慢以及求解不稳定等问题,很大程度上影响了整个反演过程的快速性和准确性。且在众多波长的消光数据中,存在较多重复冗余的信息,也很大程度上增加了反演算法的时间。针对光谱消光法粒径反演算法计算繁琐、反演效率低的问题,提出了基于主成分分析（PCA）和BP神经网络的光谱消光颗粒粒径分析方法。基于Mie散射理论对不同粒径、不同波长下的光谱消光值进行了仿真计算,通过对光谱消光数据集的主成分分析及各个波长综合载荷系数的计算,实现了最优特征波长的选取,利用降维后的光谱消光数据训练了PCA-BP神经网络模型,并利用该网络模型计算了粒径颗粒分布。通过仿真计算,比较了PCA-BP神经网络模型与传统的BP神经网络模型的预测精度,并分析了波长数目对两种神经网络模型预测结果的影响。针对训练得到的PCA-BP神经网络模型开展光谱消光法粒径参数反演算法的验证实验,搭建了光谱消光法颗粒粒径参数测量实验系统,测量了粒径范围在0.5～9.7 μm内的6种不同粒径参数的聚苯乙烯标准颗粒。仿真和实验结果表明：基于主成分分析方法可确定各个波长向量之间的相关性,利用综合载荷系数选取最优特征波长对应的消光值对整体的光谱数据具有较好的代表性,可实现光谱数据的降维。相比传统的BP神经网络模型,基于PCA-BP神经网络模型的颗粒粒径分布的分析方法预测精度更高,对于较分散颗粒系的分布参数的预测有更加明显的优势。而且,被选取的波长数较少时,PCA-BP神经网络模型依然有较高的预测精度。利用训练好的PCA-BP神经网络模型对颗粒粒径参数进行实验验证,预测结果可瞬时输出,颗粒粒径分布误差在5%以内,验证了该算法的可行性。     

Machine Learning and Micromagnetic Studies of Magnetization Switching

Magnetization switching is one of the most fundamental topics in the field of magnetism.Machine learning(ML)models of random forest(RF),support vector machine(SVM),deep neural network(DNN)methods are built and trained to classify the magnetization reversal and non-reversal cases of single-domain particle,and the classification performances are evaluated by comparison with micromagnetic simulations.The results show that the ML models have achieved great accuracy and the DNN model reaches the best area under curve(AUC)of 0.997,even with a small training dataset,and RF and SVM models have lower AUCs of 0.964 and 0.836,respectively.This work validates the potential of ML applications in studies of magnetization switching and provides the benchmark for further ML studies in magnetization switching.     

Suitable features selection for monitoring thermal condition of electrical equipment using infrared thermography

Monitoring the thermal condition of electrical equipment is necessary for maintaining the reliability of electrical system. The degradation of electrical equipment can cause excessive overheating, which can lead to the eventual failure of the equipment. Additionally, failure of equipment requires a lot of maintenance cost, manpower and can also be catastrophic- causing injuries or even deaths. Therefore, the recognition processof equipment conditions as normal and defective is an essential step towards maintaining reliability and stability of the system. The study introduces infrared thermography based condition monitoring of electrical equipment. Manual analysis of thermal image for detecting defects and classifying the status of equipment take a lot of time, efforts and can also lead to incorrect diagnosis results. An intelligent system that can separate the equipment automatically could help to overcome these problems. This paper discusses an intelligent classification system for the conditions of equipment using neural networks. Three sets of features namely first order histogram based statistical, grey level co-occurrence matrix and component based intensity features are extracted by image analysis, which are used as input data for the neural networks. The multilayered perceptron networks are trained using four different training algorithms namely Resilient back propagation, Bayesian Regulazation, Levenberg–Marquardt and Scale conjugate gradient. The experimental results show that the component based intensity features perform better compared to other two sets of features. Finally, after selecting the best features, multilayered perceptron network trained using Levenberg–Marquardt algorithm achieved the best results to classify the conditions of electrical equipment.     

A method for classifying whistles of bottlenose dolphin(Tursiops truncates) based on syntactic pattern reorganization

A method based on syntactic pattern recognition was presented to automatically classify whistles of bottlenose dolphin.Dolphin whistles have typically been characterized in terms of their instantaneous frequency as a function of time,which is also known as "whistle contour".The frequency variation features of a whistle were extracted according to its contour.Then,the frequency variation features were used for learning grammatical patterns.A whistle was classified according to grammatical pattern of its frequency variation features.The experimental results showed that the classification accuracy of the proposed method was 95%.The method can provide technical support for acoustic study of dolphins' biological behavior.     

面向智能交通系统的运动车辆阴影检测

为了解决智能交通系统中运动车辆检测鲁棒性问题,分析了阴影产生的物理机制和运动投射阴影的视觉特性,提出一种新的阴影抑制方法.该方法首先采用帧间差分法获得运动车辆轮廓和运动阴影轮廓,然后应用霍特林变换消除RGB颜色分量的相关性,最后构造了一种阴影测度来进行运动车辆和阴影的检测.实验表明该方法能有效地抑制阴影并完整地分割出运...